A new report from AI safety research organization METR, published in November 2026 following an 11-week investigation inside Anthropic, found that the company’s AI models are improving roughly twice as quickly as they were a year ago — and that AI itself is the primary driver of that acceleration.
METR’s researchers concluded that Anthropic has made “roughly two years of historical progress over the last year,” and said they are confident the speedup is not simply the result of increased computing power. Instead, the majority of the acceleration as of October is attributed to AI assistance. Anthropic employees self-reported being around 12 times as productive using the company’s most capable internal model — double what staff had reported just months earlier using a previous internal model called Mythos 5.
The report also revealed that Anthropic has completed training on three unreleased models — codenamed Model 2, Model 3, and Model 4 — finishing in July, August-September, and September-October respectively. METR found no clear signs of diminishing returns or bottlenecks that could slow further acceleration, and warned that “fully autonomous contributions will play a larger role soon.”
Anthropic was permitted to redact portions of the final report and did so extensively. METR stated that Anthropic redacted “a majority of the evidence we found most informative” from the roughly 100-page full report. Anthropic cofounder Sam McCandlish defended the redactions, saying the information was commercially sensitive and that sharing it could give rivals a roadmap to accelerate their own development.
The findings have drawn sharp criticism from AI safety researchers, who point to a contradiction between Anthropic’s public calls for responsible AI development and its apparent role in accelerating the technology. Marius Hobbhahn, CEO of safety testing firm Apollo Research, said the pace described in the report exceeds what was forecast in the influential AI 2027 scenario. Brianna Rosen of the Institute for AI Policy and Strategy called for greater government oversight, saying the report was “not sufficiently transparent about the evidence.”
New York Assemblyman Alex Bores responded to the report by calling on Congress to ban recursive self-improvement AI “this year if possible.” Anthropic declined to answer questions from The Verge, directing inquiries to McCandlish’s post on X.
Source: The Verge